Project Artifacts

Prototype Design

The focus was to create a clean, intuitive UI with minimal designs and transitions, allowing for efficient navigation
The application is designed considering users across different technological proficiency
Clear call to actions to guide the user through the UI

Ideally, this is the architecture we want to implement. The reasoning behind implementing a micro-frontend architecture is as follows:

Independent Development and Deployment:
- Each area of the application can be developed, tested and deployed individually
Scalability:
- Each module could scale individually according to the changing requirements
Tech Stack Independent:
- Each module could use a tech stack most appropriate according to it's functionality and use case
Maintainability:
- Each code base would be smaller, focused and easier to understand
- Testing would be more targeted
- Modular components would allow developers to minimize downtime

Design Document

However, for the Proof of Concept (POC), we went ahead with a monolithic implementation because of the following factors:

Development Speed and Simplicity:
- Since we were building the product out from scratch, development speed and protoyping was the focus.
- Having a single codebase makes it easier for developers to understand, test and implement the code
Deployment Efficiency:
- Since we had limited Google Cloud credits and other deployment constraints from Iowa State University' Gitlab infrastructure, we were able to create a streamlined deployment pipeline with a single Docker container, a single production environment and simple CI/CD pipeline
Resource Utilization:
- Since we had limited cloud capabilities, a microfrontend approach would mean setting up multiple smaller projects, adding to computer overhead.

Our current implementation follows a simple client-server architecture.
Frontend:
- Acts as an abstraction for all the user interactions and makes calls to the backend routes according to the functionality being used (semantic search or AI Assistant)
Backend:
- Backend receives the call from the frontend and makes necessary calls to the OpenAI API or Qdrant Database according to the query.
- Calls are made to the OpenAI API for chat completion and embedding queries
- The Qdrant Database is our vectorized database of choice, which has the vectorized content of about a 100 papers (scraped from ISU's digital repository).

current-design-doc